Closed LucaCinquini closed 11 months ago
Drew upgraded the Elasticsearch Helm chart from version 7.9.3 to 7.17.1, and was able to deploy the SPS stack onto the EKS 1.27 cluster that Tom setup on the SSIPS Test venue.
Some issues when executing tests with the 1.27 + ES 7.17.1 combination - RabbitMQ not able to connect. Ryan and Drew to work together on this.
Drew, Ryan encountered all sorts or problems when deploying SPS on 1.27 - it's possible that the EKS cluster is wrong. Previously could not pull images, it seems like it was transient. The latest issue is that some of the K8s services do not start. On the bright side, the ES service is starting up.
Latest update: APIs behind the Classic Load Balancer (WPST, SPS, JobsDB) are not reachable. Drew will try to use the latest branch from Ryan that uses a Network Load Balancer instead. He will have to manually install the K8s controller for the CLB.
Latest update: Ryan switched SPS to use a NLB instead of a CLB, after fixing a bug with the security groups now all endpoints are reachable. Also merged Drew's changes about using ES 1.27 and ES 7.10.
Ryan and Drew integrated all changes and successfully deployed SPS with upgraded ES on the new "unity-veue-dev" and on a new 1.27 cluster on "mcp-ssips-test". After switching to the Network Load Balancer, all endpoints work and were able to execute CHIRP workflow through WPS-T.
Must upgrade the Helm Chart used for Elasticsearch to a new version (7.10) because the current Elasticsearch Helm Chart requires EKS 1.24, which will be discontinued as of 10/31/2023